GitLab

When we block a rport and the driver implements the terminate
callback we will fail IO that was running quickly. However
IO that was in the scsi_device/block queue sits there until
the dev_loss_tmo fires, and this can make it look like IO is
lost because new IO will get executed but that IO stuck in
the blocked queue sits there for some time longer.

With this patch when the fast io fail tmo fires, we will
fail the blocked IO and any new IO. This patch also allows
all drivers to partially support the fast io fail tmo. If the
terminate io callback is not implemented, we will still fail blocked
IO and any new IO, so multipath can handle that.

This patch also allows the fc and iscsi classes to implement the
same behavior. The timers are just unfornately named differently.

This patch also fixes the problem where drivers were unblocking
the target in their terminate callback, which was needed for
rport removal, but for fast io fail timeout it would cause
IO to bounce arround the scsi/block layer and the LLD queuecommand.
And it for drivers that could have IO stuck but did not have
a terminate callback the unblock calls in the class will fix
them.

v2.
- fix up bit setting style to meet JamesS's pref.
- Broke out new host byte error changes to make it easier to read.
- added JamesS's ack from list.
v1
- initial patch
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Acked-by: James Smart <James.Smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

We do want to call right back into the queuecommand during the race,
so we can just use SCSI_MLQUEUE_TARGET_BUSY.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Acked-by: James Smart <James.Smart@Emulex.Com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

For the conditions below we do not want the queuecommand
function to call us right back, so return SCSI_MLQUEUE_TARGET_BUSY.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

If the fcport is not online then we do not want to block IO to all ports on
the host. We just want to stop IO on port not online, so we should be using
the SCSI_MLQUEUE_TARGET_BUSY return value.

For the case where we race with the rport memset initialization
we do not want the queuecommand to be called again so we can just use
SCSI_MLQUEUE_TARGET_BUSY for this.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Acked-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

When qla4xxx begins recovery and the iscsi class is firing up to handle
it, we need to retrn SCSI_MLQUEUE_TARGET_BUSY from the driver instead
of host busy, because the session recovery only affects the one target.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Acked-by: David C Somayajulu <david.somayajulu@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

SCSI-ml manages the queueing limits for the device and host, but
does not do so at the target level. However something something similar
can come in userful when a driver is transitioning a transport object to
the the blocked state, becuase at that time we do not want to queue
io and we do not want the queuecommand to be called again.

The patch adds code similar to the exisiting SCSI_ML_*BUSY handlers.
You can now return SCSI_MLQUEUE_TARGET_BUSY when we hit
a transport level queueing issue like the hw cannot allocate some
resource at the iscsi session/connection level, or the target has temporarily
closed or shrunk the queueing window, or if we are transitioning
to the blocked state.

bnx2i, when they rework their firmware according to netdev
developers requests, will also need to be able to limit queueing at this
level. bnx2i will hook into libiscsi, but will allocate a scsi host per
netdevice/hba, so unlike pure software iscsi/iser which is allocating
a host per session, it cannot set the scsi_host->can_queue and return
SCSI_MLQUEUE_HOST_BUSY to reflect queueing limits on the transport.

The iscsi class/driver can also set a scsi_target->can_queue value which
reflects the max commands the driver/class can support. For iscsi this
reflects the number of commands we can support for each session due to
session/connection hw limits, driver limits, and to also reflect the
session/targets's queueing window.

Changes:
v1 - initial patch.
v2 - Fix scsi_run_queue handling of multiple blocked targets.
Previously we would break from the main loop if a device was added back on
the starved list. We now run over the list and check if any target is
blocked.
v3 - Rediff for scsi-misc.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

Use a special request for serialisation purposes and get rid of the
awkward ide_spin_wait_hwgroup(). This also involves converting the
ide_devset structure so it can be shared by the /proc and the ioctl code.
Signed-off-by: Elias Oltmanns <eo@nebensachen.de>
[bart: use rq->cmd[] directly]
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

* Make ->io_buffers method return number of bytes transferred.

* Use ide_end_request() instead of idefloppy_end_request()
  in ide_floppy_io_buffers() and then move the call out to
  ide_pc_intr().

* Add ide_io_buffers() helper and convert ide-{floppy,scsi}.c
  to use it instead of ide*_io_buffers().

There should be no functional changes caused by this patch.
Acked-by: Borislav Petkov <petkovbb@gmail.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

Preparation for ide_{floppy,scsi}_io_buffers() unification.
Acked-by: Borislav Petkov <petkovbb@gmail.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

* Use drive->media and drive->scsi to check if ->dsc_overlap
  can be set by HDIO_SET_NICE ioctl in generic_ide_ioctl().

* Remove unused ->supports_dsc_overlap field from ide_driver_t.

There should be no functional changes caused by this patch.
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

* Add struct ide_devset, S_* flags, *DEVSET() & ide*_devset_*() macros.

* Add 'const struct ide_devset **settings' to ide_driver_t.

* Use 'const struct ide_devset **settings' in ide_drive_t instead of
  'struct ide_settings_s *settings'.  Then convert core code and device
  drivers to use struct ide_devset and co.:

  - device settings are no longer allocated dynamically for each device
    but instead there is an unique struct ide_devset instance per setting

  - device driver keeps the pointer to the table of pointers to its
    settings in ide_driver_t.settings

  - generic settings are kept in ide_generic_setting[]

  - ide_proc_[un]register_driver(), ide_find_setting_by_name(),
    ide_{read,write}_setting() and proc_ide_{read,write}_settings()
    are updated accordingly

  - ide*_add_settings() are removed

* Remove no longer used __ide_add_setting(), ide_add_setting(),
  __ide_remove_setting() and auto_remove_settings().

* Remove no longer used TYPE_*, SETTING_*, ide_procset_t
  and ide_settings_t.

* ->keep_settings, ->using_dma, ->unmask, ->noflush, ->dsc_overlap,
  ->nice1, ->addressing, ->wcache and ->nowerr ide_drive_t fields
  can now be bitfield flags.

While at it:

* Rename ide_find_setting_by_name() to ide_find_setting().

* Rename write_wcache() to set_wcache().

There should be no functional changes caused by this patch.
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

Call ide_proc_register_driver() in ide*_setup() (just before
ide*_add_settings() call) instead of in ->probe method.

Despite being basically a preparation for /proc/ide/hd*/settings
rework this is a nice cleanup in itself.
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

* Include <linux/ata.h> directly in <linux/ide.h>
  instead of through <linux/hdreg.h>.

* Include <linux/hdreg.h> only when needed.

Cc: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

Remove needless drive->present checks from ->probe methods
(device model takes care of that).
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

* ERR_STAT   -> ATA_ERR
* INDEX_STAT -> ATA_IDX
* ECC_STAT   -> ATA_CORR
* DRQ_STAT   -> ATA_DRQ
* SEEK_STAT  -> ATA_DSC
* WRERR_STAT -> ATA_DF
* READY_STAT -> ATA_DRDY
* BUSY_STAT  -> ATA_BUSY

* MARK_ERR   -> ATA_AMNF
* TRK0_ERR   -> ATA_TRK0NF
* ABRT_ERR   -> ATA_ABORTED
* MCR_ERR    -> ATA_MCR
* ID_ERR     -> ATA_IDNF
* MC_ERR     -> ATA_MC
* ECC_ERR    -> ATA_UNC
* ICRC_ERR   -> ATA_ICRC

* BBD_ERR    -> ATA_BBK

Also:

* ILI_ERR    -> ATAPI_ILI
* EOM_ERR    -> ATAPI_EOM
* LFS_ERR    -> ATAPI_LFS

* CD         -> ATAPI_COD
* IO         -> ATAPI_IO
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

* Use ATA_CMD_* defines instead of WIN_* ones.

While at it:

* EXABYTE_ENABLE_NEXT -> ATA_EXABYTE_ENABLE_NEST

* SETFEATURES_{EN,DIS}_WCACHE -> SETFEATURES_WC_{ON,OFF}

* SETFEATURES_{EN,DIS}_AAM -> SETFEATURES_AAM_{ON,OFF}

* SMART_* -> ATA_SMART_*

* Remove stale comment from ide-proc.c.

Partially based on earlier work by Chris Wedgwood.
Acked-by: Chris Wedgwood <cw@f00f.org>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

Make drive->id an unnamed union so id can be accessed either by using
'u16 *id' or 'struct hd_driveid *driveid'.  Then convert all existing
drive->id users accordingly (using 'u16 *id' when possible).

This is an intermediate step to make ide 'struct hd_driveid'-free.

While at it:

- Add missing KERN_CONTs in it821x.c.

- Use ATA_ID_WORDS and ATA_ID_*_LEN defines.

- Remove unnecessary checks for drive->id.

- s/drive_table/table/ in ide_in_drive_list().

- Cleanup ide_config_drive_speed() a bit.

- s/drive1/dev1/ & s/drive0/dev0/ in ide_undecoded_slave().

v2:
Fix typo in drivers/ide/ppc/pmac.c. (From Stephen Rothwell)

There should be no functional changes caused by this patch.
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

blk_rq_unmap_user in sg_finish_rem_req can take care of all the cases.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

sg_read_xfer was used to copy data to user space for READ
commands. blk_rq_unmap_user does the job so sg_read_xfer does nothing
useful.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

sg_write_xfer was used to copy data from user space for WRITE
commands. blk_rq_map_user_iov and blk_rq_map_user do the job so
sg_write_xfer does nothing useful.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

Calling blk_rq_map_user() at a single place is better than at
different two places. It makes the code more understandable.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

__sg_start_req() was used temporarily to call blk_get_request() during
converting sg to use the block layer.

Now sg always calls blk_get_request() so we can move blk_get_request()
to sg_start_req(). We don't need __sg_start_req anymore.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

It's not used for anything useful after the block layer conversion.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

sg had lots of the own functions for the direct IO but now sg uses the
block layer functions for it. There are only five lines for the direct
IO. SG_ALLOW_DIO_CODE define was used to compile out the direct IO
code but we don't need the define. If someone wants to remove the
direct IO code, he can do easily without the define.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

old sg_rq_end_io() was used to wrap sg_cmd_done during converting sg
to use the block layer (in order to cover the difference
scsi_execute_async and blk_execute_rq_nowait). Now we don't need it so
let's remove it.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

Right now SCSI and others do their own command timeout handling.
Move those bits to the block layer.

Instead of having a timer per command, we try to be a bit more clever
and simply have one per-queue. This avoids the overhead of having to
tear down and setup a timer for each command, so it will result in a lot
less timer fiddling.
Signed-off-by: Mike Anderson <andmike@linux.vnet.ibm.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

Modify the SCSI disk driver to call the revalidate_disk()
wrapper. This allows us to do some housekeeping such as accounting for
a disk being resized online. The wrapper will call
sd_revalidate_disk() at the appropriate time.
Signed-off-by: Andrew Patterson <andrew.patterson@hp.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

With the older SG interface, we don't know a user-space address to
trasfer data when executing a SCSI command. So we can't pass a
user-space address to blk_rq_map_user.

This patch fixes sg to pass a NULL user-space address to
blk_rq_map_user so that it just sets up a request and bios with page
frames propely without data transfer.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Douglas Gilbert <dougg@torque.net>
Cc: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

This patch converts the indirect IO path (including mmap IO and old
struct sg_header) to use the block layer functions (blk_get_request,
blk_execute_rq_nowait, blk_rq_map_user, etc) instead of
scsi_execute_async().

[Jens: fixed compile error with SCSI logging enabled]
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Douglas Gilbert <dougg@torque.net>
Cc: Mike Christie <michaelc@cs.wisc.edu>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

This patch converts the direct IO path (SG_FLAG_DIRECT_IO) to use the
block layer functions (blk_get_request, blk_execute_rq_nowait,
blk_rq_map_user, etc) instead of scsi_execute_async().
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Douglas Gilbert <dougg@torque.net>
Cc: Mike Christie <michaelc@cs.wisc.edu>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

This patch converts the non data path to use the block layer functions
(blk_get_request, blk_execute_rq_nowait, etc) instead of uses
scsi_execute_async().
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Douglas Gilbert <dougg@torque.net>
Cc: Mike Christie <michaelc@cs.wisc.edu>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

This patch introduces struct rq_map_data to enable bio_copy_use_iov()
use reserved pages.

Currently, bio_copy_user_iov allocates bounce pages but
drivers/scsi/sg.c wants to allocate pages by itself and use
them. struct rq_map_data can be used to pass allocated pages to
bio_copy_user_iov.

The current users of bio_copy_user_iov simply passes NULL (they don't
want to use pre-allocated pages).
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Jens Axboe <jens.axboe@oracle.com>
Cc: Douglas Gilbert <dougg@torque.net>
Cc: Mike Christie <michaelc@cs.wisc.edu>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

Currently, blk_rq_map_user and blk_rq_map_user_iov always do
GFP_KERNEL allocation.

This adds gfp_mask argument to blk_rq_map_user and blk_rq_map_user_iov
so sg can use it (sg always does GFP_ATOMIC allocation).
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Douglas Gilbert <dougg@torque.net>
Cc: Mike Christie <michaelc@cs.wisc.edu>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

Now that disk and partition handlings are mostly unified, it's easy to
allow disk to have extended device number.  This patch makes
add_disk() use extended device number if disk->minors is zero.  Both
sd and ide-disk are updated to use this.

* sd_format_disk_name() is implemented which can generically determine
  the drive name.  This removes disk number restriction stemming from
  limited device names.

* If sd index goes over SD_MAX_DISKS (which can be increased now BTW),
  sd simply doesn't initialize minors letting block layer choose
  extended device number.

* If CONFIG_DEBUG_EXT_DEVT is set, both sd and ide-disk always set
  minors to 0 and use extended device numbers.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

With previous changes, it's meaningless to limit the number of
partitions.  Replace @ext_minors with GENHD_FL_EXT_DEVT such that
setting the flag allows the disk to have maximum number of allowed
partitions (only limited by the number of entries in parsed_partitions
as determined by MAX_PART constant).

This kills not-too-pretty alloc_disk_ext[_node]() functions and makes
@minors parameter to alloc_disk[_node]() unnecessary.  The parameter
is left alone to avoid disturbing the users.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

Extended devt introduces non-contiguos device numbers.  This patch
implements a debug option which forces most devt allocations to be
from the extended area and spreads them out.  This is enabled by
default if DEBUG_KERNEL is set and achieves...

1. Detects code paths in kernel or userland which expect predetermined
   consecutive device numbers.

2. When something goes wrong, avoid corruption as adding to the minor
   of earlier partition won't lead to the wrong but valid device.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

Update sd and ide-disk such that they can take advantage of extended
minors.

ide-disk already has 64 minors per device and currently doesn't use
extended minors although after this patch it can be turned on by
simply tweaking constants.

sd only had 16 minors per device causing problems on certain peculiar
configurations.  This patch lifts the restriction and enables it to
use upto 64 minors.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

* Implement disk_devt() and part_devt() and use them to directly
  access devt instead of computing it from ->major and ->first_minor.

  Note that all references to ->major and ->first_minor outside of
  block layer is used to determine devt of the disk (the part0) and as
  ->major and ->first_minor will continue to represent devt for the
  disk, converting these users aren't strictly necessary.  However,
  convert them for consistency.

* Implement disk_max_parts() to avoid directly deferencing
  genhd->minors.

* Update bdget_disk() such that it doesn't assume consecutive minor
  space.

* Move devt computation from register_disk() to add_disk() and make it
  the only one (all other usages use the initially determined value).

These changes clean up the code and will help disk->part dereference
fix and extended block device numbers.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

On Mon, 2008-09-22 at 14:56 -0700, akpm@linux-foundation.org wrote:
> From: Andrew Morton <akpm@linux-foundation.org>
>
> s390:
>
> drivers/scsi/device_handler/scsi_dh_emc.c: In function 'parse_sp_info_reply':
> drivers/scsi/device_handler/scsi_dh_emc.c:179: warning: comparison is always false due to limited range of data type
>
> because chars are unsigned, I assume.

Fix by making csdev->buffer explicitly an unsigned char and dropping
the < 0 test.
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>